Go to Statistics Portal

VARIANCE ESTIMATION

Statistics Directorate    
Definition:
The task of estimating the value of the variance of an estimate.

Context:
Methods employed are commonly classified as:

- analytic methods: use and compute the proper formulae of the variance;

- approximate methods: methods that use approximations for complex and muti-stage sample designs.

These methods can be classified in three categories:

- simplifying assumptions: setting assumptions which allow use of straightforward formulae (as if an analytic method could be applied).
- Taylor's linearisation techniques: estimation of non-linear statistics is simplified in using a Taylor's development of the concerned statistics.
- Replication or re-sampling methods: a large number of sub-samples are derived from the initial sample and variance is estimated from the variability of the results in this set of sub-samples. Several methods are based on re-sampling, such as the jack-knife, the bootstrap and the balanced repeated replication methods.

Source Publication:
Eurostat, “Assessment of Quality in Statistics: Glossary”, Working Group, Luxembourg, October 2003.

Cross References:
Estimate
Variance

Statistical Theme: Methodological information (metadata)

Created on Friday, November 8, 2002

Last updated on Wednesday, January 4, 2006